Specifications

DATA CENTER BEST PRACTICES
SAN Design and Best Practices 38 of 84
When using a four-box solution, it is inappropriate to make ISL cross-connections between the two channel
extenders within a data center site and both the “A” and “B” FC fabrics, because of the reasons discussed
above. However, it is permissible to do so on the Ethernet/WAN side (see Figure 25).
fig26_SAN_Design
Production Site Back-Up Site
A
FC
FC
FC
FC
B
A
B
Brocade
7800
Storage
Array
Storage
Array
FCIP
Brocade
7800
Brocade
7800
Brocade
7800
WAN
IP Network
Figure 25. Ethernet connectivity to dual-core WAN infrastructure.
FCIP with FCR
The FCIP tunnel traditionally traverses a WAN or IP cloud, which can have characteristics that adversely impact
a Fibre Channel network. The FCIP link across a WAN is essentially an FC ISL over an IP link. In any design,
it should be considered an FC ISL. Repeated apping of a WAN connection can cause disruption in directly
connected fabrics. This disruption many come about from the many fabric services trying to reconverge, and
reconverge again, and reconverge again, over and over. This causes the CPU on the switch or director to max
out. If the CPU can no longer process the various tasks required to operate a fabric, there may be an outage.
If you limit the fabric services to within the local fabric itself and do not allow them to span across the WAN,
you can prevent this from occurring. FCR provides a termination point for fabric services, referred to as a
“demarcation point.” EX_Ports and VEX_Ports are demarcation points in which fabric services are terminated,
forming the “edge” to the fabric. A fabric isolated in such a way is referred to as an “edge fabric.” There is a
special case in which the edge fabric includes the WAN link because a VEX_Port was used; this type of edge
fabric is referred to as a “remote edge fabric.”
FCR does not need to be used unless there is a production fabric that must be isolated from WAN outages.
When connecting to array ports directly for RDR, FCR provides no benet. Mainframe environments are
precluded from using FCR, as it is not supported with FICON.
Please note, when a mainframe host writes to a volume on the Direct Access Storage Device (DASD), and that
DASD performs RDR to another DASD, then DASD to DASD trafc is not using FICON. It is an open systems
RDR application such as EMC SRDF, HDS HUR or TrueCopy, IBM Metro Mirror or Global Mirror, or HP Continuous
Access. These open-system RDR applications can use FCR, even though the volumes they are replicating were
written by the FICON host.
There are some basic FCR architectures:
•First and simplest, no FCR or one big fabric: this type of architecture is used with the mainframe and when the
channel extenders are directly connected to the storage arrays.
•Second, edge-backbone-edge, in which edge fabrics bookend a transit backbone between them.
•Third, when a VEX_Port is used, the resulting architecture can be either backbone-remote edge or edge-
backbone-remote edge, depending on whether devices are connected directly to the backbone or an edge
fabric hangs off of the backbone. Both are possible.